AD-A236 350 .

RL-TR-91-45Final Technical ReportMay 1991

5/15 GHZ SCATTERING STUDY

Thayer School of Engineering

Robert K. Crane DTICMýECTE

APPROV FOR PUJC RELEASE, DO1SrU1naVST'-AWIM.T9

91.0068 4

Rome Laboratory I11Air Force Systems Commai -

Griffiss Air Force Base, NY 13441-5700

This report has been reviewed by the Rome Laboratory Public AffairsDivision (PA) and is releasable to the National Technical InformationService (NTIS). At NTIS it will be releasable to the general public,including foreign nations.

RL-TR-91-45 has been reviewed and is approved for publication.

APPROVED: 64e /, L/, ,ýr"-

UVE H.W. LAMMERSProject Engineer

APPROVED:

jORN K. SCHINDLERDirector of Llectromagnetics

F'OR THE COMMANDEK". 6

J••ES W. HYDE TIIh)irectoraLe of Plans & Programs

If your address h2, changed or if you wish to be removed from the RomeLaboratory mailing list, or if the addressee is no longer employed by yourorganization, pleas# notify Rome Laboratory (EECP) Hanscom XFB MA01731-5000. This will assist us ti maintaining a current mailing list.

Do not return conies of this report uni.ess contractual obligations ornotices on a specific document require that it be returned.

REPORT DOCUMENTATION PAGE FOrm AoproveNPkh rk n@M g ithfnz • e* d - o to.wbn. is I tomq I •ouar peruis t g t nm fh tt dfr*M i n n Wb u ig Q kig dKa swc• .galw -dn'•qttt doln Irmid o " wtid ~ • as- w*q znkt•Itro d 0 kft m Siot ammuts-lod I q ttdin bd NIoTn or uN w aslpet of t•Scokamct, e' ~TWanum r.&x* aopabnet frck, mlaq " hsft W80*19 1`1441#Aft SVrtK 01 H for kftawum Oppatk andRqectta, 121 S Jeffrson,DeH aOg'uW. SL* 12r4, Ai* iVA2r.4a-=•l a OfIbsd M i10 andStui PouW1( AedX.R Pdpd (MO .I. WOW&UM CC 20=

1. AGENCY USE ONLY (Leave Blanh) 2. REPORT DATE 3.REPORT AYPE AND DAT'E$ COVERF,ay 1991 Final Jun 88 - Aug 89

4. TITLE AND SUBTITLE 5. FUNDING NUMBERS

5/15 GHZ SCATTERING STUDYI C - F30602-88-D-0027Task E-8-7109

PE - 28010F, 63789F6. AUT1HOR) PR - 478T

& jb..rt K. Crane TA - 00WiU - P1

7PERF. AMINt! ORGANZATION NAME(S) AND ADDRESS(ES) a PERFORMING ORGANtZATON

Thayer ScI'ool of Engineering REPORT NUMBE.RDartmouth College N[ L.,nover NH 03755 N/A

C.. ,1PONkOiNGQMON[TORiNG AGENCY NAME(S) AND ADORES(ES) 10, SP0MN•RONG/MONITORING

Roan Laboratory (EECP) AGENCY REPOFT NUMUER

Hanscom AFB MA 01731-5000 RL-TR-91-45

11. SUPPLEMENTARY NOTES

Rome Laboratory Project Engineer: Uve H.W. Lammers/IEEC/(6 17) 377-3193

Prime Contractor: Syracuse University

12a, LSTR1BU11ONJAVARABUTY STATEMENT 12bL DSSIvRAON CCDU

Approved for public release; distribution unlimited.

Received signal level and multipath delay spread measurements were made it 5 and 16

CHz on a 161 km tropospheric scatter path over moderately rough terrain from Prosp(cct

Hill1 in Waltham, MA to Ht Tug in Lehanon, N1l. The measurement campaign spanned thesummer monthv of May through Aug 89. The signal level data were processed to obtain

hourly median values for the estimation of the cumulative distribution function (cdf)

of received signal level for use in troposcatter communication syste" design. Multi-

path delay spread observation,4 were made at the higher frequency. These data were

also processed to obtain the sampl cdt of the hourly median values of the two s

Block 13 (Cont'd)

NOTE: Rome Laboratory/RL (formerly Rome Air ))evelopment Center/RA.DC

Abstract

Rec, ved signal level and multipati; delay spread measurements were made at 5 and 16GHz on a 161 km tropospheric scatter path over moderately rough terrain from Prospect Hill inWaltham Massachusetts to Mt Tug in Lebanon, New Hampshire. The measurement campaignspanned the summer months of May through August, 1989. The signal level data wereprocessed to obtain hourly median values for the estimation of the cumulative distributionfunction (cdf) of received signal level for use in troposcatter communication system design.Multipath delay spread observations were made at the higher frequency. These data were alsoprocessed to obtain the sample cdf of the hourly median values of the two sigma multipathdelay spread estimates

The data were sorted by th- dominant propagation mechanism for each hour ofobservation. The propagation cataories were clear weather conditions (forward scatter byclear air turbulence in thip horizontal layers), rain scatter, and duct propagation in elevatedlayers. The sample cdfs were well approximated by lognormal distributions for received signallevel and by normal distributions for multipath delay spread. At 16 GHz (Ku band), thehighest level field strengths were recorded during periods with rain while, at C band (5 GHz),the highest level fields were recorded during elevated ducting conditions.

r-L

Accession forNNS VRA&DTIC 7aB 0Un,,ulourncod 0Juit irl eat!on

By_-Ditr buttroseAvailability Codes

.ýn Rka1 an'd/r

-t

Table of Contents

Abstract " i

Table of Contents ii

List of Tables ii

List of Illustrations iii

Acknowledgements vi

1. Introduction I

1.1 Program objectives 1

1.2 Summary of results 1

2. The experimental observations 3

2.1 Equipment configuration 3

2.2 Daily observations with the small aperture Ku band transmit antenna 4

2.3 Observations with the large aperture Ku band transmit antenna 7

2.4 Observations in rain 8

2.5 Observations with an elevated duct 8

3. Data summaries 10

3.1 Hourly data 10

3.2 Diurnal variations 11

3.3 Statistical distributions of hourly median values for clear weather conditions 12

3.4 Statistical distributions of hourly median values for rainy conditions 14

3.5 Statistical distributions of hourly median values for ducting conditions 15

4. Analysis 16

References 17

List of Tables

I. Quarter day summaries.

List of Illustrations Page

1. Troposcatter path profile for a 4/3 effective Earth radius. The scattering volumes are 20displayed as filled areas. The scattering volume for the 3' antenna at 16 GHz enclosesthe scattering volume for the 29' antenna at 5 GHz.

2. Day plots for the small aperture Ku band transmit antenna under clear weather conditions 21on July 12, 1989: sample cumulative distribution functions for received power, receivedpower time series, power spectra for received power fluctuations and coherency betweenthe fluctuations at Ku and C band.

3. Day plots for the small aperture Ku band transmit antenna under clear weather conditions 22on July 12, 1989: time series of received power deviations from turbulent forward scattermodel predictions, time series of the differences between received power deviations fromturbulent forward scatter model predictions at Ku and C band, time series and samplecumulative distribution function for Ku band two sigma multipath delay spread.

4. Day plots for the small aperture Ku band transmit antenna under clear weather conditions 23on July 12, 1989: time series of mean Doppler frequency shift and Doppler frequencyspread, time series tod sample cumulative distribution function for carrier channel anddelay channel measurements of Ku band received power levels..

5. Day plots for the laige aperture Ku band transmit antenna under clear weather conditions 24on July 12, 1989: sample cumulative distribution functions for received power, receivedpower time series, power spectra for received power fluctuations and coherency betweenthe fluctuations at Ku and C band.

6. Day plots for the large aperture Ku band transmit antenna. under clear weather conditions 25on July 12, 1989: time series of received power deviations from turbulent forward scattermodel predictions, time series of the differences between received power deviations fromturbulent forward scatter model predictions at Ku and C band, time series and samplecumulative distribution function for Ku band two sigma multipath delay spread.

7. Day plots for the large aperture Ku band transmit antenna under clear weather conditions 26on July 12, 1989: time series of mean Doppler frequency shift and Doppler frequencyspread, time series and sample cumulative distribution function for carrer channel anddelay channel mneasurenwes of Ku band received power levels.

8. Scatter distribution comparing simultaneous observations of the maximum hourly carrier 27"channel and delay channel observations of Ku band received power levels.

9. Day plots for the small aperture Ku band transmit antenna under rainy conditions on July 285. 1989: sample cumulativ- distribution functions for received powr, received powertime series, power spectra for received power fluctuations and coherncy between thefluctuations at Ku and C band.

10. Day plots for the small aperture Ku band transmit antenna under rainy conditions on July 295. 1989: time series of received power deviations from turbulent forward scatter modelpredictions, time series of the differences between received power deviations fromturbulent forward scatter model predictions at Ku and C band, time series and samplecumulative distribution function for Ku band two sigma multipath delay spread.

II. Day plots for the small aperture Ku band transmit antenna under rainy conditions on July 305. 1989: time series of mean Doppler frequency shift and Doppler frequency spread,, time

WU

series and sample cumulative distribution function for carrier channel and delay channelmeasurements of Ku band received power levels.

12. Day plots for-the small aperture Ku band transmit antenna under elevated ducting 31conditions on August 1, 1989: sample cumulative distribution functions for receivedpower, received power time series, power spectra for received power fluctuations andcoherency between the fluctuations at Ku and C band.

13. Day plots for the small aperture Ku band transmit antenna under elevated ducting 32conditions on August 1, 1989: time series of received power deviations from turbulentforward scatter model predictions, time series of the differences between received powerdeviations from turbulent forward scatter model predictions at Ku and C band, time seriesand sample cumulative distribution function for Ku band two sigma multipath delayspread.

14. Day plots for the small aperture Ku band transmit antenna under elevated ducting 33conditions on August 1, 1989: time series of mean Doppler frequency shift and Dopplerfrequency spread, time series and sample cumulative distribution function for carrierchannel and delay channel measurements of Ku band received power levels.

15. C band carrier received power levels for the month of June, 1989. 34

16. Ku band carrier received power levels for the month of June, 1989. 35

17. Ku band delay channel received power levels for the month of June, 1989. 36

18. Ku band average Doppler frequency shifts for the month of June, 1989. 37

19. C band average Doppler frequency shifts for the month of June, 1989. 38

20. Ku band Doppler frequency spreads for the month of June, 1989. 39

21. C band Doppler frequency spreads for the month of June, 1989. 40

22. Ku ba.nd two sigma multipath delay spreaids for the month of June, 1989. 4'1

23. Average diurnal vanation in C band received arrier power, swummr. 1989 -12

24, Average diurnal variation in Ku band roceived carrier power. sumnner, 1989 4'3

n Average diurnal variation in C band Doppler frequency shift. swunmer. 1989 44

26, Average diurnal variation in Ku band Doppler frequency shift. sunroer, 1989 45

27. Average diurnal variation in Ku band Doppler spread. swnme, 1989 -16

28. Average diurnal variation in Ku band two sigma multipath delay spread. stmurt . 1989 47

29. Sample cumulative distributions for carrier received power levels, nighttime. summer, 481989. clear weather conditions.

3'0 Samplc cumulative distributions for carrier received power levels, morning. sum-e,.1989. clear wv:ther conditions.

iv

31 Sample cumulatve distributions for carrier received power levels, afternoon, summer, 501989, clear weather conditions.

32 Sample cumulative distributions for carrier received power levels, evening, summer, 511989, clear weather conditions.

33 Sample cumulative distributions for carrier received power levels, large aperture antenna 52at Ku band, afternoon, summer, 1989, clear weather conditions.

34 Sample cumulative distributions for Ku band Doppler spread values, summer, 1989, 53clear weather conditions.

35 Sample cumulative distributions for Ku band two sigma multipath delay spread values, 54summer, 1989, clear weather conditions.

36 Sample cumulative distributions for carrier received power levels, nighttime, summer, 551989, rainy -onditions.

37 Sample cumulative distributions for carrier received power levels, morning, summer, 561989, rainy conditions.

38 Sample cumulative distributions for carrier received power levels, afternoon, summer, 571989, rainy conditions.

39 Sample cumulative distributions for carrier received power levels, evening, summer, 581989, rainy conditions.

40 Sample cumulative distributions for Ku band Doppler spread values, summer, 1989. 59rainy conditions.

41 Sample cumulative distributions for Ku band two sigma multipath delay spread values, 60summer, 1989. rainy conditions.

42 Sample cumulative distributions for carrier received power levels, nighttime. sunmmer. 611989, elevated ducting conditions.

43 Sample cumulative distributions for canier re.,ived power lrvels. morimng. sun-UeIr. 6219M(9. elevated ducting conditions.

44 Sample cumulative distributions for Ku band Doppler spread values, sunmier, 1989. 63elevated dtutiag conditions.

45 Sample cumulative distributions for Ku band two sigma multipath delay spread values, .summeur. 1949. elcvated ducting conditions.

46 Sample cumulative distribuious for carrier received powvr lcvels for difekrent 65p gaoiao n echanisum .unamer. 1989.

V

Acknowledgements

The experinfient described in this report was sponsored by the Joint TacticalCommunications (TRI-TAC) Program Office, Electronics Systems Division (ESD) ard theRome Air Development Center (RADC). The work at the Thayer School of EngineerinG.,Dartmouth College was supported by ESD and RADC through the RADC Post-Doctol...Program. Uve H. W. Lammers and Richard A. Marr of the Electromagrietics Directorate,Rome Air Development Center, Hanscom Air Force Base, MA, were responsible for theoperation of the transmitter site.

The receiver site on Mt. Tug is the cable head facility for Twin State Cable of WestLebanon, NH. The generosity of Twin State Cable in allowing us to use the facility isappreciated.

A number of graduate students at the Thayer School of Engineering have been involvedin this project at one time or another. Special thanks go to Brad Anderson and Mark Hoppe forthe effort they have expended in collecting Lnd analyzing the data. Thanks also go ShaileshChandra and Rajesh Prathu for picking up !.he work and continuing the observation andanalysis progiam after the graduations of Brad and Mark.

1v

5 / 15 GHZ SCATrERING STUDY

1. Introduction -

1. 1 Progrant ubjectives

The objectiv ! -f this 5/15 GHz Sc,.,:;,ing Study was to characterize the 16 GHztroposcattcr cliant,) for the c,,, ,f '•igital tactical communication systems. Simultaneousmeasurements of -eceived sigiz-. !.YeL%. -.-, ,..ade at two frequencies, 5 and 16 GHz, togetherwith measuremc:nts of the Doppler frequency shtft 4f the received carrier signals azid themultipath delay spread of the higher frequency sgnal. The parameters needed to assess systemperformance, the received signal level, delay spread and Doppler spread (fade rate), werederived from the measurements. factical troposcatter communication systems presentlyoperate at C band (5 GHz) and, therefore, the lower frequency observations were made toprovide a known reterence for comparison to the higher frequency (Ku band) measurements.

An earlier study had shown that adequate signal levels were present on the 161 kmProspect Hill to Mt, Tug troposcatter path when a large aperture antenna was used at thetransmitter site [Crane, 19881. Questions were raised about the signal levels to be expectedwhen small aperture antennas were employed at both the transmit and receive sites. This studyalso addressed the effect of antenna aperture size on received signal level statistics, delayspread statistics and Doppler spread statistics.

1.2 Summary of results

The transmitter and receiver systems for the Prospect Hill to Mt Tug troposcatter pathwere configured for continuous, unattended operation with small aperture, matched be•anwidthantennas at the receiver site for both frequencies and at both ends of the path for tle higherfrequency. The digital data recording system collected measurements of the carrier receivedpower levels at 4.95 GHz (C band) and 15.73 GHz (Ku band), the average Doppler fvequencyshift and the standard deviation of the Doppler shift at both carrier frequencies and, at thehigher frequency, the cerrelation detector received power levels at 15 consecutive lags (80nanosecond time intervals) straddling the delay with the highest detected level. T1e receiversystem was calibrated automatically twice a day. The observations were processed to providehourly summaries that included the minimum, maximum and nmedian values of two minuteaverages of each of the measured paramctern. These sun•a•ics wer' then stored in a datahank foýr subs•equent analyses.

Data were collectei from I May through 21 August 1989. Operation was nearlycontinuous with the exception of a daily shutdown of otie to two hours for maintenance andcalibration and a series of aftcmrnoo measurements during July with the large aperture tranmUtantenna at Ku band. Problems with one of the two Ku band tran: hitters caused the loss of%even days of data in May eight days in June and thirteen days in July. Uhe problem wasfixed by the end of July but operations terminated August 21 when all three transmitter. faileda% the r :sult of a lightning strike at the tmnsmnitter site. During the four nv'mth period up to thelightning strike. calibrated observations were obtained for 1232 hours W. 4*% of the tinm.

The observattons revealer a wtrong diurnal variation in received signal level. Theobserved average hourly median tran smssion loss values were 163 dB at Ku band and 1 49 dBat C band for time penods with no indication of rain on the path and no enhanced signal levelsdue to propagation along elevated ducts. The corresponding Ku band signal level was withinI-' dR of the predictions of several troposcztter •-,odels. The Signauon model [Parl. 1999:Matthews. 19891 predicted aNbut 8 dB more signal than observed. The current CCIR momklICCIR. 19Q•9 Wredicted 12 dB more signal than observed while their earlier modee, the N'83

Tech Note 101 model [Rice et. al., 19651, predicted 3 dB less signal than observed. Our mostrecent model is within 3 dB of the average value for the summer months. At C band, themodel prediction errors were smaller.

The average and spread (standard deviation) of the Doppler shifts of the carrierfrequencies were monitored throughout the experiment. The average Doppler shift was used toidentify intervals with rain scattered signals. The Doppler spread may be used to estimate thefade rate for the received signals. At Ku band, the observed Doppler spread averaged 8.6 Hz,a value close to &e predicted value of 8.1 Hz if the prevailing mind is 12 m/s across the path atthe height of the scattering volume. At C band, the observed Doppler spread was 2.2 Hzwhile, tor the same average wind conditions as at Ku band, the Doppler spread should havebeen 1.0 Hz (corresponding to 90 fades per minute). The residual frequency variations in thetransmit and receive equipment were too large to make successful Doppler spreadmeasurements at C band and at all but the highest spread values at Ku band.

The average hourly median two sigma delay sprea,, was 138 nsec. This value is identicalto the predicted output from the RAKE correlation receiver system for a 12.5 Mbit/s bit rate i ."typical receiver signal-to-noise ratios if the scattering is dominated by a thin (100 m)r'satteri.-. ;layer at a 1 km height. For the small aperture antennas and a smooth, slowly varying profi!, ofscattering from clear air turbulence (uniformly filled scattering volume), the expected delayspread increases to 174 nsec. The observations were comparable to the 150 to 160 nsecmedian delay spread measurements made on a 138 km path with W antennas at C band using a10 Mbit/s RAKE system ISherwood and Suyemoto, 19761. The theoretical calculations ofdelay spread indicate that the actual spread values should be less than 40'., of the measuredvalues (for a receiver with infinite bandwidth and no receiver noise).

Measurements were made using the large, 29' aperture antenna at Ku band at thetransmitter site during twelve afternoons in July. The C band system was unchanged and wasemployed to provide the reference for the performance comparison between the large and smallaperture antennas. The large antenna had an estimated gain of 58 db at Ku band which is 16dBl higher than the estimated gain of the small, 3' antenna. Model calculations predict a hourlymedian transmission loss for !he large iperture antenna that is only 2 dB lower than thepredicted hourly median transmission loss for the smaller antenna for a uniformly filledscattering volume or 3.5 dB lower for a thin scattering layer. After normalization so the Cband measurements had identical average hourly median transmission loss values for thesummer aftermoon time period (equal to 145 dB). the average nicasured hourly ncdiantransmission loss was less than I dB lower than expected for a unifonnly filled scatteringvolumie (at 156 dB) ft• the large antenna white the transmission toss, was less than I dB higherthan expected for ;,he small antenna (at 1514 dB). The ohsrrved differences evwenobservatttn and predtction for either thin layers or a unifoimly filled sattering v•lunie were%-61 w!th1t tzhe iceasumrnu errors of the C and Ku banld tr'vni•u•s• syyt•-ms.

For the la,•er apcrturt antenna, the predicted [X-pler spread and delay spread valu•• a&resnaller than for the setall antenna. For the summter aftcrnotn data ets. the othcr'ed !ok'lerspread at Ku hand for the large antenna was half the value measured for the small antenna 13 9% ,7 77 t W. A% prtdtcted. the average hiurly mcdian multipath Jelav sfread was als& smalterfor the Lujge antenna t' 9 v% I 'S nmc).

lVork twrfornied under this 5ubeontract has also been reprined in the master's theses:"T0hervatno arnd inaly•i% of rr ,sphcnc satter grgopgavton dunng rami at 16 (st1 W.a 569lW" hy Bradley T A.nderson 1 1991 and The Study of the Ku-biand Tv 'K-pcric Scatter(hannel Dunng C(ear. ar Co-dwons t•,.a Data Co•!ected On the Link Rerstueen Psi P t Hill inWaltham. MA amd Ndt Tug Ln Fnfield. Nli- ty Mark A ltrp[ I VIF

2. The experimental observations

2.1 Equipment configuration

The earlier measurement campaign [Crane, 19881 employed the large, 29' apertureantenna at Prospect Hill in Waltham, Massachusetts for transmission at both frequencies. Thereceiver site on Mt Tug in Lebanon, New Hampshire had both standard gain horns andmatched beamwidth parabolic reflector antennas that could be used for reception. The hornswere employed for calibration and antenna pointing checks; the matched beamwidth antennaswere utilized for data collection. As originally configured, the receiver system could use eitherthe C or Ku band carrier frequencies for the phase reference for coherent detection Thesignals were mixed down to baseband in the receivers, low pass filtered and sampled (buh in-phase and quadrature components) for subsequent computer processing. The final low passfilters had a 120 Hz bandwidth.

The receiver system was modified to use the cesium beam frequency standards (one at thetransmitter site and one at the receiver) as the phase reference for coherent measurements. Theabsolute Doppler frequency shifts could then be observed instead of the relative Doppler shiftbetween the two carrier frequencies when one or the other carrier frequency was used toprovide the phase reference. The data processing software was upgraded to provide Dopplershift and Doppler spread estimates based on the pulse-pair algorithm and the complexcorrelations between consecutive samples of the in-phase and quadrature signals.

The multipath delay profile measurements were made using correlation detection on a1023 bit pseudo-noise sequence generated by a maximum-length shift register code (PRNcode). Identical code sequences were produced at the transmitter and receiver. The transmittersequence was phase modulated on the 15.73 GHz carrier at a 12.5 Mbit/s bit rate. The codesequence generated at the receiver was used to phase modulate one of the local oscillators.Correlation detection was accomplished by mixing the modulated local oscillator with the downconverted received signal and averaging in the low pass filters (in-phase and quadrature).Under computer control, the receiver code could be shifted forward or backward in timerelative to the transmitted sequence to change the time lag for correlation detection. Thecomputer program automatically selected the lag giving the greatest output and, for datacollection, shifted the sequence ± 7 lags to straddle the peak. For each lag, 200 samples of thethe in-phase and quadrature channel output were simultaneously recorded at a 250 Hz rate foreach channel.

The multipath delay spread algorithm was upgraded to obtain a continuous measurementof receiver noise level and to subtract the estimated noise power froin the correlation detectoroutputs at the different lags. Two sigma delay spread estimates were calculated from the noisecon'ected power vs delay profile and averaged for two minutes prior to final recording.Received power cstimates were calculated from the sum of the recorded powers for eachcorrelation detector output (each lag) and from the carrier channel output. The correlationdetector channel hd an 18 dB higher signal-to-noise ratio than tne carrier channel thusproviding superior performance at low signal levels.

The received power should be spread over only a limited number of lags when turbulentlayer scdtter is the dominant propagation mechanism. When rain scattering occurred on thepath, the delay spread was expe¢,ted to increase dramatically. Initially, the position of the lagwindow used for the delay spread measurements was adjusted from one 30 secondmeasurement cycle to the next. The anticipated slow drift of the cesium standards relative toeach other could then be corrected and unattended delay spread measurement would bepossible. LUnfortunately, during rain the larger delay spread: caused the tracking system to"lose lock" and further delay measurement was generally not possible until lock was manuallyreacquired. The correlation detector power estimate was used to keep track of the operation of

3

the tracking system but could not be relied upon for routine. power measurement. When it wasfound that the cesium beam standards did not drift fast enough rel~ative to each other to requirecontinuous tracking, the algorithm was chainged to disable tracking after lock was initiallyacquired. The system worked reasonably well after that with lock being lost only occasionally.Delay spread measurements were then possible in rain.

A new Ku band traveling wave tube (TWT) transmitter was inm;talled at Prospect Hill foruse with the small, 3'aperture antenna. The older klystron transmitter that was used with thelarge, 29' antenna required continuous supervision while the new system could be rununattended. The C band TWT transmitter was also used with the large aperture antenna but it",ould also be run unattended. A possible drawback to the. new Ku band system w-as its lowertransmit power. Model calculations predicted that the use of 3.5 dB less transmit power -dthe expected 2 to 4 dB increase in transmission loss (decrease in signal level) caused by.echange from the large to small aperture transmit antenna would niot significantly affect themeasurements. The expected median signal-to-noise ratio was still better than 20 dBl with thenew transmitter system and, for th.- correlation detection channel, the expected signal-to-noiseratio was better than 38 dB.

The Prospect Hill to Mt Tug troposcatter path profile is displayed in Figure 1. The figurepresents a cross section view of the path in the great circle plane. The scattering volumes forthe several antenna configurations are displayed as shaded areas. The scattering volumes arebounded by the 3 dB contours of the antenna patterns and the radio horizon rays from eachantenna. The antennas were pointed at their local honzons along the great circle path. Thereceiver antenna beamwirdths were mratched but the transmiitter antenna beamwidths varied withaperture size and frequency. Thie resulting common volumes (scattering volumes common !oboth antenna pa- -'ns) were long, horizontally oriented cylinders. At Ku band, the commonv.olume was 69 t.:l when the small aperture transmit antenna was used but was only is~kmi long for the large aperture antenna. At C band, the length of the common volume was 36kmi. The lower edges of the comnmon volumes were 820 mn above mean sea level and 400 mabove the terrain. Tiie maximum vertical extents of the common volumes were I060 m for thesmrall antenna at KU band, 2-40 m for the large antenna at the same frequency and 520 m at Chand. The m~xiOlium crosts-path horizontal extents were twice the vertical extents. Althoughthe common volumes were long and thin, scattering was possible anywhere in the volume thatwas line-of-sight it) the anteiinas at each end of the path. Rain scattering was important %shentne rain cells occurred over either antenna witnin the main lobe of one antenna and the far sidelobes of the other as well as when the cell appeare-d within the common volume.

The trrmin profile aloNg the great circle path was relaively ffin from the transmitter site oabout 70) kmn from that site. The remainder of the path was hilly with a height variation of 40X)m. A succession of wooded hills broke uip the path. At least 4 diffracting obstacle ridgesoccurred on the path. The diffraction loss for this path significantly exceeded the maximumexpxzetd tranismission loss for a turbulent volume scatuer path. The path was asymmetrical

Aith a higher horizon angle at the receiver than at the transmitter. 1lie center% of thi scaitenogvolumeis were there-fore closer to the receiver site.

2.2 Daily observationis with the small aperture K u band transmit antenna.

The observations wert recorded continuously at Jir receiver site. For each paramietci.2W( samples were gzathered in a 0.8 second interval for processing to estimate the averagevaluc. The C band and Ku band carrier channels were sampled simultaneously hut, for thedifferent lags of the Ku ban~d correlation detection channel, the samples were obtainedsequentially. A full cycle for detecing the signal levels and setting the attenuators ito place thesignal in thec ccntcr 'of the dynamic range for each receiver channel, for Sam1pling all the,:hanncls. and for calculatiE'g the parameter estimates and recording their values took 30secconds. The Im'ear. coherent rcceivers had limited dynamic ranges and digitally controlled

4

attenuators were required to keep the widely varying signal within the dynamic range. Thecalculations included computations of the received power levels (non coherent processing) andof the mean and standard deviation of the Doppler frequency shifts (coherent processing).

Four consecutive observations were averaged for display and further analysis. Figures 2through 4 present a full sequence of output for a single day of measurements with the smallapertuie antenna at Ku band and the normal system at C band. The upper rOght hand panel inFigure 2 presents the received signal levels from the carrier channels. Each recorded twominute average is displayed. The upper trace is for the C band signal level and the lower traceis for Ku band. The signal levels have been adjusted (calibrated) to represent the output fromthe receiver for a 30 watt (w) transmit power level. The actual Ku band carrier signal level atthe input to the transmit antenna was 3.6 w. for C band the transmit power was 30 w. Thereceiver sigral levels were referenced to the input port of the waveguide switch that selects thereceive antenna or calibration noise diode for connection to the receiver. Transmission loss isdefined to be the measured power level difference (loss) be:ween the transmitter and receiverwaveguide reference points. The receiver noise levels (after adjustment to compensate fordifferences in transmit power relative to the 30 w reference) are also displayed in the figure.

The upper ieft hanai panel in the figure displays the sample cumulative distributionfunctions (cdfs) for the received signal levels at each frequency. The abscissa for the cdfs isthe reduced variate for a normal probability distribution. If the cdf follows a straight line in thisplot, it could be approximated by a lognormal distribution (the ordinate is in dB). Both cdfsappear to follow straight lines for the fraction of the day with data.

The lower panels present the signal level fluctuation power spectra and the coherencybetween the fluctuations at each carrier frequency, The power spectra were computed from thetime series of two minute averages. The spectra are averages of consecutive 32 sample spectra,apprcximately one hour duration). The two minute averaging was done to reduce anyvariations due to Rayleigh fading on the troposcatter channel, spectral averaging was used toreduce the statistical uncertainties of the spectral estimates. The resulting spectra represent theslow variations in ihe parameters of the turbulent process that produced the fading. The semi-empirical st.ustical theory for turbulence in the clear atmosphere predicts that the spectra shouldobey a f 5 (f = fluctuation i requency) power-law relationship. This relationship is displayedas a dot-dashed line. The Ku band spectrum follows the expected relationship for fluctuationfrequencies below 0.001 Hz. The C band spectrum shows relatively more higher frequencyfluctuations than predicted. The coherency between the fluctuations at the two carrierfrequencies is essentially zero. The differences in spectral shape and lack of coherency suggestthat the turbulent layers or patches are either thin enough or of small enough horizontal extentto affect ote but not both scattering paths (with different common volume sizes).

Figure 3 presents the relative behavior of the scattering channels (upper panels) and theto sigma multipath delay spread observations at Ku band. For the upper left hand panel, thereceived signal level deviations from the predictions of our model for forward scattering byclcar air turbulence are displayed for both carrier frcquencies. If the intensity of scattering fromthe turbulence was distributed in space in accordance with the model. the received signal traceswould lie along the horizontal. 0 dB line (dot-dashed). If the turbulence had a uniformintensity throughout both common volumes, the traces for both canter frequencies would lie ontop of each cher. The observations show differences in the positions of the curves of as mucha1 12 d13. These differences are plotted in the upper right hand panel. Usually, for scatteringby turbulence the deviations between the model normalized curves are less than 6 dB.

The two sigma multipath delay spread closely followed a normal distribution (lower righthand panel) with a median spr.ad value of 140 nsec. The predicted spread values varybetwecn 140 nsc for a thin. 1()0 m layer and 173 nsec for turbulence filling the commonv.olumnc I after correction for system bandwidth and receiver noise). For a point, intense

scatterer (infinite signal-to-noise ratio, zero delay spread) the expected two sigma delay spreadvo' lues vary from 0 to 80 nsec depending on the delay of the scattered signal relav'.e to the starttime for a lag bin. Averaging over a uniform distribution of possible relative start times, ast.-ong point scaierer would produce a 62 nsec two sigma multipath delay spread value Thecdf f.r delay spread shows that the scatterers produce more spread than a single point scatt,;rerand generally less spread than expected for a uniformly filled scattering volume.

The left hand panels in Figure 4 summarize the Doppler frequency shift measurements forthe day. The expected mean Doppler shift is zero. The turbulent scatterers move horizontallyand, for a scattering path that is symmetric about the great circle path, the mean Doppler shiftshould be zero. A close examination of the local horizon at the receiver site shows that ther.ointiiig direction to produce the smailest scattering angle is 0.2 degrees to the east of the greatcircle direction. Because the scattered signal levels are a streng function of the scattering angle,the path is not symmetrical about the great circle plane and the mean Doppler shift should varywith the horizontal wind at the height of the lower edge of the scattering volume. For theprevailing air movement from the west, the result is a small negative average Doppler shift.Because the lower edge of the scattering volume is shaped by the local horizon and not the;nterina beams, the magnitude of the observed Doppler shifts should vary in proportion to thecarrier frequency. For time periods when the model normalized received signal levels coincide(0400 to 0800 h local time) the Doppler shifts were in the ratio of the carrier frequencies. Atother times, the mean Doppler shifts were not simply related. Several occurrences of large,short period (single two minute sample) Doppler shifts are evident in the data. These areattributed to szatteri'ig by ,iircraft. The aircraft has to fly almost parallel to the scatter path for asufficient tin- period to produce an output that survives the 120 Hz narrow hand filtering andsubsequent averaging over the 9.8 second sampling interval and the two minute (4 sample)averaging interval. The commuter flights from Boston, Massachusetts to Lebanon, NewHampshire have flight paths that can produce the obsered scattering signatures.

The Doppler shift data were prima-ily used to identify periods with rain on the scatterpath. For rain i , the common volume, the expected Doppler shifts are greater than +9 Hz atKu ban-d and +3 Hz at C band. The expected rain scatter signatures would be clearly evident inthe data in the figure.

The Doppler spraa estimates are presented in the lower left hand panel of the figure.The obse"vations at Ku band show some meteorologicai variation but the C bandmeasurements do not. The Doppier spread is produced by the horizontal motions of thescatterers througnuut the common volime. ý 'jdel calculations show that the observed Kuband spread values are consist-Mi with cross path winds of the order of 10 rn/seC at a 1 kmheight. The C hand observations am• more than twice the value expected for such winds. TheDoppler spread estimation algorithm is very c:nsitive to the receiver signal-to-noise ratio andestimates were r.ot made for very low signal-to-noise values. T"he spread estimates m,'ny also becontaminated by low frequency phase mrdulation of the trnnsmitted catTier or any of thereceiver local oscillators. The C band trw.. ,iitter had a 60)Hz modulation (later found toonrginate from a f~r;urd loop) that contributed to the large residual Doppler spread values. TheKu bhand TWVT also had relaively strong 60. 9"i and 120 Hz modulation line,; but ihe pi.as.,hilfts due to aunospheric motions were still dc•ninant

The right. tand pariels in Pigure 4 display the performance of the correlation detectionchannel rceative to tr'c carrier channel. If all is working properly and the delay spread is smallenough to keep the multipath delay profile within the lag window used for nrocessing, the Kucarrier and Ku delay signal levels should be identical. The upper right hand panel displaysboth received signal level estimates and the estimated receiver nozie ievel (normalized to thereceived signal for a modulated transmitted signal power of 30 w ic. 9 dB below the actualreceived noise power and 18 d3 below the value reported for the carrier signal levelob.ervations in Figures 2 and 3). The cdf shows the median delay channel power e.timate is 2

6

dB higher than the median carrier power estimate for all periods when both channels wereoperating. The delay channel power leveis varied about the carrier channel values by as muchas 3 dB. The variations in the differences between the carrier and delay channel observationsare attributed to statistical sampling errors. For an 0.8 second sample of a Rayleigh fadingprocess with a fade rate of 1.5 times th, Doppler spread, approximately 10 independentsamples were used for the estimation of each received power level value (30 second samples).The expected deviations between the two minute averages of the power level estimates for eachfrequency should then exceed 3.8 dB for 10 percent of the samples. The varying differencesbetween the two curves are therefore consistent with the expected statistical uncertainties in thepower level estimates for each channel. For intervals with smaller Doppler spreads, thestatistical uncertainties of the measurements will be larger. The difference in the median valuesis larger than expected by chance and indicates a calibration error in either of the receiverchannels or between the total power measurements at the transmitter and the expected residualcarrier leakage through the balanced modulator that is used for the carrier signal for detection.

"2.3 Observations with the large aperture Ku band transmit antenna

Data from the time interval between 1000 to 1700 h is missing from Figures 2 through 4.During this interval, the large aperture antenna was employed for transmission at bothfrequencies . The observations with t,,e large aperture antenna are presented in Figures 5through 7. In these figures, the C band measurements are as reported above. The Ku bandmeasurements were also processed as above but the calibration constants were for the smallaperture antenna and TWT transmitter and not for the large aperture antenna and klystrontransmitter. As a result, the reported Ku band observations are 8 dB higher than they shouldbe. The difference in calibration is due to the differences in transmitter power, klystronbandwidth and the level of the residual carrier level leakage through the modulator.

"The shapes of the cdfs and power spectra curves are nearly the same at the twofrequencies and the coherency between the signal level fluctuations is high at the lowerfluctuation frequencies. The C band data show a marked increase in signal level (-15 dB)during the 1000 to 1200 h interval. The Ku band carrier signal followed the C band carriersignal variations within the measurement (sampling) uncertair,'s for each channel.

The power levels from the correlation detector channel track the carrier power levels forKu delay channel output values below -105 dBm (as reported with the calibration constants forthe small aperture antenna transmitter). Above that level, the dynamic range of the receiversystem was exceeded, the delay channel saturated, and the delay channel receiver noise levelincreased in response to receiver saturation (Figure 7). Normally, the receivers are operatedwith sufficient IF gain to keep the aioise levels in the narrow band stages of the carrier channelhigh enough to mask the slowly varying de blases produced in the final stage of mixing downto baseband. The high intermediate IF gains did not provide a sufficient dynamic range (40 dBatxwe the expected levels in the carrier channel) to accommxlate the higher level signals whenthe large aperture transmitter was in use. Figure 8 display,; the maximum carrier channel signallevel and delay channel signal level values for each hour with simultaneous data. The effect ofreceiver saturation is clearly e-vidtnt (the data for the large aperture ar.tenna transmissions islabeled July LA). By the end of July when the small aperture transmitter system was again onthe air, the receiver gains and automatic gain control algorithm were adjusted to eliminate thesaturatlon pr9oblem (but not the dc bias problem).

The two sigma multipath delay spreads were normally distributed and perhaps smallerthan for the small aperture antenna. Figure 6 does not contain sufficient data to make acomparison. The effect of receiver saturation should be small because the noise subtractiotistep in the multipath delay spread estimation algorithm used the receiver noise estimatescalculated for each measurement cyCle. Saturation would reduce the received power values atthe lag having the highest signal but. for the large aperture antenna, saturation would affect

7

only two lags at most with the result that the delay spreads estimates, which are alreadyseriously over estimated, would have only slightly more error.

The mean Doppler shift measurements appear to be a continuation of the small apertureantenna observations but insufficient data are available in the figure for a comparison betweenthe obse•'vadons at different aperture sizes. The Doppler spread values are significantly smallerthan for the small aperture transmit antenna. Model calculations suggest that the observedvalues are consistent with the same 12 m/s cross wind velocity invoked to explain theobservations with the small aperture antenna. ThI observations show smaller hour-to-hourvariations than expected or observed with the small aperture system. The problem may be thatthe small aperture results contain both meteorological and equipment produced spreadcomponents, the former time varying and the latter constant in time. With the expected smallercontributions to the meteorologically varying component, the large aperture antenna results maybe dominated by equipment produced phase noise.

2.4 Observations in rain

Figures 9 through 11 illustrate the effects of rain on the troposcatter signal for a day ofobservations using the small aperture antenna transmitter system. The set of three figuresprovide the ;ame information as before. The carrier signals experienced more hour-to-hourvariations than for the clear air day (Figures 2 through 7). The shorter time scale variationswere reduced in magnitude and the coherency between the fluctuations was small. The dailymedians of the C band and Ku band received signal levels were about 10 dB higher than for theclear air day. The cdf for the C band received signal levels was nearly lognormal bu', for Kuband, the signal levels show a 10 dB or more decrease below the lognormal distribution for thesmallest 5% o. ihe observations (1.6 standard deviation and higher).

The mean Doppler shift record (upper left hand panel in Figure 11) show markedincreases in the Doppler shift from 0500 h through the end of the day. This is the signature ofrain on the path. Scattering by rain dominates over scattering by clear air turbulence at Kuband over a wide range of rain intensities while, at C band, it is dominant for only very highrain rates. The result is a rain signature at Ku band but no evidence of rain at the lower carrierfrequency. The relative magnitudes of the Ku and C band signals show higher signals at Kuthan at C band relative to the turbulent model predictions during most of the intervals with largeDoppler shifts. An exception is evident between 2200 and 2300 h. During this short interval,the Ku band signal fell to more than 20 dB below the level expected from the C bandweasuretents for scattering by turbulence. The problem is attenuation by rain on the path."The scattered signals are stronger than for clear air turbulence but attenuation by rain betweenthe antennas and the scattering volume can reduce the signal levels by more 'han they areenhanced by raiw scatter.

During the rain attenuation event, the carrier signal is reduced to near the noise levelcausing, in part, the increased Doppler spread observed during the event. The multipath delayspread increased to m~re than 4,X) nscc and exceeded the lag window used for Doppler spreadanalysis. Although some of the power was spread outside the analysis window, sufficientremained inside the window to show that the attenuation was not large enough to cause a losso(f signal. During the periods with rain on the path the delay spread values were consistentlylarger than for times with clear air scatter.

2.5 Obscrvationis with an elevated duct

The third propagation mechanism thought to be important on a troposcatter path acrossmoderately rough terrain (height variations of the order of 400 m) is ducting or trapping in anelc',ated layer (or duct). The terrain is too rough to support surface ducts and, therefore, that

8

mechanism can be ignored. Also, the path has too many diffraction ridges for high level fieldsto occur due to diffraction during super refraction conditions.

The identifying features for coupling via atmospheric ducts are enhanced signal levels,especially at the lower frequency, combined with a significant reduction in the fade rate [Crane,19811. Because the ducting process is caused by refractive bending and trapping within anatmospheric waveguide, for a thick enough waveguide the phase variations observed on thepath are not produced by a Doppler shift upon scattering but by the much slower variations inthe integrated refractive index along ,he ray paths and in the physical lengths of the multiple raypaths between the temporally and spatially varying top and bottom of the guide. As a result,the average Doppler shifts should be reduced to near zero and show little variation during theintervals with slow changes in signal strength.

Figures 12 through 14 show intervals of elevated layer ducting between 0000 and 0800 hand again from 2200 h through the end of the day. The C band signal levels are more than 20dB stronger than the Ku band levels and more than 10 dB larger than expected for scattering byclear air turbulence at either frequency. The fading rate is slow enough to be evident in the timeseries of two minute averages and the mean Doppler shift at C band is essentially zero. A shortperiod of enhanced signal levels was observed at Ku band from 0100 to 0200 h. During thistime interval, the Ku band Doppler shift was zero and the delay spread was twice reduced to avalue approaching the spread value expected for a point scatterer or a strong path with noadditional multipath components (such as for a line-of-sight path).

9

3 Data summaries

3.1 Hourly data -

The time series of two minute averages were partitioned into clock hour blocks (eachstarting on the hour). For each full clock hour of observations, the maximum, minimum, andmedian values were obtained and stored in a computer data base for further analysis. Each ofthe important variables: carrier received signal levels, Doppler shifts, and Doppler spreads wererecorded as were the delay channel received signal level, two sigma multipath delay spread,and receiver noise level.

The dominant propagation mechanism for each bour was established automatically. Fivepropagation conditions were identified: clear air, raw, elevated ducting, rain attenuation andaircraft scatter. Most of the observations were designated as corresponding to clear airconditions. Hours were identified as dominated by rain scatter if both the maximum andmedian Ku band Doppler shifts were larger than preset thresholds (3 lIz for the median and 6Hz for the maximum). If only the maximum Doppler shift exceeded the threshold and themedian value was less than zero, the hour was identified as dominated by aircraft scatter.Hourly intervals with median and maximum C band signal levels significantly larger than theKu band levels and with near zero Doppler shift at C band were identified as dominated byducting in elevated layers (the thresholds were 17 dB for the median difference and more than20 for the maximum values). The fifth category was rain attenuation. In this case the hourlymedian Ku band signal was more than 27 dB below the expected value for clear air turbulenceas calculated from the C band measurements (more than 12 dB net reduction in signal level)and rain scattering was identified in that or an an adjacent hour interval. A final manual checkwas made of the identified propagation mechanisms to insure that the rare occurrences ofelevated duct propagation at Ku band would not cause the interval to be excluded from theducting category.

The hourly value time series for the month of June are presented in Figures 15 through22. The received carrier levels at C band are displayed in Figure 15. Three curves are plotted,the minimum and maximum hourly observations are the thin traces while the median value isgiven by the thicker trace between the thin traces. The curves are continuous wheremeasurements were available for contiguous hours. The blanks indicate missing data. Themissing observations are generally for periods when the Ku band transmitter failed. Afternoonobservations on several of the days late in June were excluded because transmission wasswitched between the large and small antenna systems at half hour intervals. Receiver andtransmitter system calibrations were then hard to maintain and the switched measurementcampaign was abandoned for the series of afternoon observations with the large antenna thatwas conducted in July.

The propagation mechanism code is indicated by the positions of the diamond markersacross the top of the figure: at -78 dBlm for rain, -76 dBmn for rain attenuation and -80 dBm forelevated ducting conditions. No markers are displayed for either clear air turbulence scatterconditions or for aircraft scatter. Elevated ducting conditions occurred during the nighttime.early morning hours on the 20th through the 23rd of June.

Figures 16 and 17 present the Ku carrier and delay channel received power levelsrespectively. The periods of enhanced signal levels with elevated ducting is evident in thecarrier channel observations but not in the delay channel output where receiver satur'ation hastaken its toll. The lowest Ku band signal levels observed during the summer were recordedduring 15 - 16 June. Most of these observations were not associated with a rainy period. Thesignal levels were relatively low at both C band and Ku band although the Ku band signals areabout 5 dB below the levels predicted from the C band measurements for scattering by clear air

10

turbulence uniformly fil, g the common volumes. The signal levels were high enough to bedetected in the delay chai..,el with about a 20 dB signal-to-noise ratio. The noise level is givenby the thick curve along the bottom of Figure 17. The receiver noise level variations are not anindication of changes in receiver gain (and calibration) but are produced by the slowly varyingdc biases generated in the final stage of mixing down to baseband.

Figures 18 and 19 display the median and extreme hourly values of the average Dopplershift for the Ku and C band carriers respectively. The periods with rain are clearly identified inthe Ku band data. These data also show the slow meteorological changes in the horizontalwind in the scattering volume between rain events. The sharp, isolated positive and negativeDoppler deviations are indicative of aircraft scatter. During the extended interval with lowsignal levels at Ku band, the average Ku band Doppler frequency dropped to -6 Hz. Thetransmitted carrier frequency is offset from the receiver local oscillator frequency by 6 Hz whenmixed down to baseband. Correction for the frequency offset is made in the calculation ofDoppler shift. The average frequency for noise is 0 Hz as is the frequency of the dc biasesproduced in the firal mixing stage. After compensation, a zero frequency output from thereceiver is reported as a -6 Hz Doppler shift. During the low signal level interval, the Dopplershift estimates were contaminated by the dc biases and receiver noise.

Figures 20 and 21 display the hourly Doppler spread values. At C band, the meanDoppler shifts showed significantly smaller deviations from zero than for Ku band. "Te shiftsat the two frequencies were generally of the same sign indicating that both channels wereresponding to the motions of the scatterers. Both the C band and Ku band Doppler spreadobservations show isolated incidents of zero values. These arise in the spread estimationalgorithm when a rapid change in carrier frequency occurs such as can be produced byscattering from aircraft.

Figure 22 presents the multipath delay spread observations for the month of June. Ingeneral, large values of spread are associated with occurrences of rain, small values with theelevated ducting occurrence and intermediate values for scattering by turbulence in the clearatmosphere. The magnitudes of the delay spread values during periods of clear air conditionsindicate that the turbulence is confined to layers that are substantially thinner than the verticalextent of the common volume.

3.2 Diurnal variations

The observations for June showud day-to-day and within-e-day variations in signalstrength. The hourly median data for the entire summer were separated by propagationmechanism, sorted by time of day, and averaged to display the diurnal variations and relativemagnitudes of the different propagation mechanisms. Hours dominated by aircraft scatter werenot included in the analysis because too few observations were available; the rain attenuationevents were included with the rain events. Only 7 of the 1232 hours of observation weredesignated as dominated by rain attenuation (0.57% or 3.2% of the time with rain)

Figures 23 and 24 depict the diurnal variations of the received signal levels at the twocarrier frequencies. The clear weather condition data show an avernge received signal level ofabout -105 dBm at C band (corresponding to a transmission loss of 150 dB) during theevening and early morning hours. At about 1000 h, the signal levels increase, reach a peakvalue of about -97 dBm tan increase of 8 dB) at about noon, then decline in magnitude for therest of the day. This cycle also occurs at Ku band with a nighttime level of about -120 dBm(165 dB transmission loss) and the daytime peak at -113 dBm. Turbulence in the planeta.yboundary layer increases in intensity in response to warming by the sun. The thickness of theboundary layer grows during the morning hours and by 1000 h the top of the layer crosses thelower edge of the cornmen volumes.

11

Rain can occur at anytime of the day or night. In contrast to the diurnal variations in th"intensity of scattering by turbulence, the observed changes in the average signal levels duringperiods with rain show only a decrease in the late evening that may not be statisticallysignificant.

Coupling via an elevated duct occurs only at night. The signal levels at C band are attheir highest during the early morning hours. For the observations during the summer 1989campaign, the signal levels were more than 10 dB higher than for scattering by clear airturbulence.

The average Doppler shift values show diurnal changes in response to the changes in thepropagation mechanisms. In Figures 25 and 26, the average Dopplec shifts are smaller duringthe daytime period when boundary layer turbulence occurs in the common volume and arehigher when the scattering comes from the stable region above the boundary layer. Underducting conditions, the the Doppler shifts are the smallest. During rain, the Doppler shifts arelargest in response to scattering by the falling raindrops. Under clear weather conditions, theobserved average Doppler shifts are proportional to carrier frequency. They are more than anorder of magnitude larger than expected for scatterirg volumes that are symmetric about thegreat circle path. The problem is the asymmetry forced by the variations in the elevation anglesto the radio horizon with azimuth from the great circle path at the receiver.

The Doppler spread values presented in Figure 27 were larger at night than during theday. The spread values were about the same for periods with elevated ducting and for clearweather conditions. This results from the relatively smail increase in Ku band signal level incomparison to the change at C band when ducting occurs. At Ku band, rain scatter producedhigher Doppler spread observations.

The average delay spread values ai"o increased during periods of rain as shown in Figure28.

3.3 Statistical distributions of hourly median values for clear weather conditions

The hourly observations were grouped by quarter d&y intervals for the preparation ofsample cumulative distributions. The following time intervals were selected for analysis:nighttime, 2200 to 0400 h; morning, 04,0 to 1000 h; afternoon, 1000 to 1600 h; evening,16(W) to 2200 h. Referring to Figure 23, the intervals were selected to group contiguous hourswith similar signal levels. For clear weather conditions, the nighttime period corresponds to thelowest levels: the afternoon period corresponds to the highest levels. Table I presents theaverages and standard deviations for each measurement by quarter day interval and bypropagation mechanism.

The sample cumulative distribution finctions (cdfs) for the observed median hourlyreceived carrier power levels are presented by quarter day intervals for clear weather conditionsin Figures 29 through 32. The thick curves are the sample cdfs. The scale for the reducedvariate for a nonnal distribution is the number of standard deviations of that distribution. Thedot-dashed lines art the median t.xpected cdfs for lognonral distributions (the ordinate has alogarithmic scale) having an average value (lognormani m in Table 1) near the observed averagevalue at C band and a model predicted value at Ku band based onl the C bnd value. The modelwas for forward scattcring by clear air turbulence. The standard deviations for t',e lognonlaldistributions (lognormal s in Table 1) were picked to match #.he slopes of the C band and Kuband distributions. The dashed curves above and below each lognormn nodel distributionbound the expected 5% to 95% range for sample cdfs obtained from the observed number ofsamples (assumed to '.v independent of each other) from the model probability distribution. If

12

the sample cdf was constructed from observations drawn from the model distribution, it wouldlie between the dashed curves with probability 0.9.

The lognormal model provides an excellent approximation to the C band signal levelsample cdfs for all the quarter day intervals. The separation between the bounding curvesdepends upon the standard deviation of the assumed lognormal distribution and the number ofindependent samples used for the construction of the sample cdf: the fewer independentsamples, the wider the bounds. The bounds were calculated on ýhe basis of the number ofhours of data listed in Table 1. The lognormal model was combined with the forward scattermodel for clear air turbulence to predict the parameters of the lognormal distribution at Kuband. The forward scatter model provides two predictions depending upon the assumedstructure of the turbulent region within the common volumes, 15.2 dB lower signal strengths atKu band than at C band if the turbulence is entirely enclosed within a layer (or layers) that arethin in comparison with the vertical extents of both common volumes and 13.5 dB lowersignal strengths at Ku band if the turbulence ".iiformly fills both common volumes. Inpractice, the ratio should lie between the two values. The value used for the calculation of theKu band lognormal m value is listed in Table I as the lognormal m value in the column labeled"Ratio of Carrier Levels". The iognormal model for Ku band also provides an excellent fit to!he sample cdfs.

The standard deviation values (lognormal s) varied little with time of day, from 5.5 dBfor the nighttime and morning periods, to 5.8 dB in the afternoon, and to 5.0 dB in theevening. At Ku band, a larger variation but of smaller values was required to provide a goodmatch to the data, between 3.5 and 5.2 dB. The Ku band values were consistently smaller thanthe C band values. This result is in agreement with the predictions of the empirical modelcurrently suggested for use by the CCIR [19891 for transhorizon propagation paths. TheCCIR model attributes the reduction in variance to the increase in carrier frequency. Thatmodel cannot be correct because, for the large aperture antenna, the standard deviation valueneeded to match the lognormal model to tne sample cdf the standard deviation increases withfrequency not decreases. The physically based turbulent volume scattering model depends onfrequiency through an explicit model for the change in the bistatic scattering cross section peruiit vlume wikh fre.quency and an implicit model for the change in scattering volume witharitemin- size and sc itr path geometry. The explicit model for frequency dependence affectsonly the .- edian :.-gaat levels. 'he standard deviation values depend only on the relative sizesof Che scattering volumes. Our observations show 'he !arger the scattering volume, the smallerthe variance. The expectation is that for scattep% 0y tutrbulence, the variances would beidentical if the scattering volumes were matched at the different freque1 cies.

The sample cdfs for observations with the large aperture antenna are presented in -igý -":33. Thie thick, solid curves are the sample cdfs, the thin dot-dashed curves are the expectcdmedian distributions for the lognormal models with the parameters listed in table 1. Thedashed curves are the expected 5%. 95% bounds for the sample cdfs. As above, the averagevalue for the lognormal model at Ku band was calculated from the value at C band. Turbulentstatter model calculatiorns predict a 11.5 dB difference between the C baud and Ku bant carrierpo•we- levels when the scattering volume for the large antenna is filled at Ku band. The thin,gray. solid curves are from the median lognonnal model distributions for the small apertureantenna measurements reported in Figure 3 1. The C band system was unchanged for the twomeasurement sets and was recorded to provide a reference for the comparison between thelarge and small transmit antenna at Ku band. For the 12 days of observations with the largeantenna, the C hand signal levels were higher on average than for the rest of the 4 month datast. At the median level for the cdfs (0 reduced variate), the expected 5% to 95% ranges for thecdfs overlap showing statical agreement between the afternoon observations on the differentmeasurement days. The Ku band observations are consistently higher for the large antennathan for the small antenna and overlap occu..: only at the low signal level tail of the distribution1.The sample cdf at Ku hand is in excellent agreement with the lognornial model predictions

13

5/15 Scattering Study

showing that the apparent differences between the sample cdfs for the large and small antennaswere expected. The standard deviation value needed to fit the tails of the distribution as well asthe central region is larger for the large antenna at Ku band than for C band (and for the smallantenna at Ku band).

The Doppler spread at Ku band provides information about the fade rate to be expected atthe higher carrier frequency. Comparison is not possible to the spread values at C bandbecause equipment noise was significantly larger than the , ariations due to propagationconditions on the path. Equipment induced frequency variations were also important at Kuband. The results are shown in Figure 34. In this case, only the high Doppler spread tail ofthe distribution shows any response to variations caused by conditions on the path. Theobservations could be used to set an uppei limit to the spread values to be expected or, the path.A better approach would be to estimate the cross path wind velocities in at the height of thecommon volumes and calculate the Doppler spread using the scattering model. In that way.more than the upper tail of the distribution would be used to estimate the parameters for thedistribution.

The delay spread is another parameter that is needed for communication system design.The sample cumulative distributions for each quarter day interval are displayed in Figure 35.Because little variation was observed with time of day, the four sample cdfs were plotted on asingle graph. The sample cdfs are well approximated by normal distributions. A modeldistribution for the number of samples in the afternoon time period was fit to the data. Themorning and evening observations are consistent with the model distribution. The model forthe nighttime data should have a higher average delay spread and a higher standard deviationvalue. For the afternoon time period, the average and standard deviation values should belower. The ranges of the model parameters are small as illustrated in the column of Table Ilabeled "Ku band Delay Spread".

3.4 Statistical distributions of hourly median values for rainy conditions

Sample cdfs were :onstructed for rainy conditions on the path for each of the qnarter dayintervals. The received signal level distributions are displayed in Figures 36 through 39. Themodel distributions plotted in eac .i figure are for clear weather conditions (Figures 29 through32) but with the model bounds calculated for the number of hours of observations used for theconstruction of the sample cdfs. They are included for reference. in Figure 36, the C bandmeasurements lie within the expected bounds of the lognormal model fit to the clear weather Cband observations for the same time period. At Ku band the observations depart from thebounds about the nxtel fit to the clear weather data. Rain scattering .nroduces a significantincrease in signal level relative to clear weather conditions for more than 60)% of the time %% ithrain. The signal levels are smaller than expected during clear weather conditions for less than15% of the time with rain. The smaller signal levels are presumably duO tn roni at:.nuati:'., :-ithe path havin-g more of an effect than the inceasc due to alattenrng.

For t.,: morning hours. Figure 37, the signal levels arc higher in rain than iM. the clear atbNth frequencies. At Ku band the enhancement in signal level is more pronounced than at Chand. For the afternomn time pcn

3.5 Statistical distributions of hourly median values for ducting conditions

Figures 42 through 45 present the sample cdfs for times with occurrences of enhanceusignal levels at C band due to trapping in elevated ducts. Figures 42 and 43 present the carrierpower level cdfs for the nighttime and morning hours. Enhanced signal levels relative to theexpected clear weather, turbuent laye: scatter levels occur at both trequencies. The signallevels at C band are signiticanmly higher as required by the conditions set to identify a timeinterval as associated with ducting condition.. The additional requirement of a small averageDoppler shift (magnitude less than 0.5 Hz) limited the selection such that the signal leveldistributions for clear weather and ducting conditions overlapped in signal magnitude.

The Ku band delay and Doppler spread disuibutions do not deviate significantly from theclear weather sample cdfs. Deviations from the Doppler spread model distribution are evidentfor all the propagation models but this is caused by the equipment.

15

4. Analysis

The propagation mechanism for each clock hour of data from the '.umrner 1989measurement camnpaign was identified on the basis of the signal level and average Doppler shiftobservations. With the -eparation into the three basic categories used for analysis, clearweather, rain and elevated ducting conditions, the sample cdfs for received signal level for eachcategory could bh modeled by lognoraal probability distributions. The san.ple cd!., forrLiltipath delay spread could ec modeled by normal probability distributions. Th,: lognr, mialhehavior for the received pomer level distributions was evident in the quarter day data. Thedifferences between the median value parameters (lognormal m) from one quarter day intervalto the next were of the same magnitude as the standard deviation values for a tiElre intervAl Thesingle composite sample c~dfs for all the observations for a single propagation mechanism w,,erealso lognorrmal. They are presented in Figure 46. This figure presents a suniniury of the signallevel observations for the entire surmoer.

The hourly median clear weather received carrier power observation,: are consistent withthe lognonnal nmxoel for signal levels at both C band and Ku band. The median values for thesummer months are - 103.5 IBm (148.3 dB transmission loss) at 4.95 G1lz and -118.4 dBm

163.1 dB transmission loss) at 15.73 Gl-z. The standard deviations are 6 dB at 4.95 C3Hzand 5 dB at 15.73 GHz. The sample cdfs lie entirely within the expected 5% and 95% boundsfor lognonnal cdfs with the median and standard deviation values listed above. They areconsistent %k ith the hypothesis that each hourly median value is an indepenident sample fr-wn alognomial prxcess.

For rain, the cdf of the hourly median received carrier power levels at Ku band isconsistent with a lognormal model distribution with a median value and standard deviationlarer than for clear weather conditions: -115.1 dBrn ( 159.9 dB transmission loss) and 7.1 dIMrespectively. At C hand, the rainy condition obsevations do not differ significantly frolm theclear weather conditions except for the highest 2% of the obse-ed values.

During pventds with coupling via elevated ducts, the median and standard deviation,.lues for the hourly median received carer power levels at C band are significantly higherthan for clear weather conditions: -91.3 Minm t136t1 dii transmission loss) and 9.0 d(repcctlvely. The ndiatin level and Standard deviation values at Ku haad are comparable to theKu hand values for rain. Tlhe sample cdt" at C band is also cnmstcnt with the lognoirmaldstribution hypothesis.

The ratio% of the rcceived carrier poawcr levels am the twvi' frequencies are consisttt withthe predtictions of a thin laver, turbulent volumc forward %catter m•oel for t*th the large andsmall ap(rture transnit antennas at Ku band and the large aperture trarsmit antenna at C band..A\% predicted. the signal levels thit result fruttn the use of the small vitenna at 16 (l1, were not%,gnificantly riri ,r than the svgnal% obtained using the larg:e aplr'•rr antenna. Tkhe •s.at rnddl.,A hen used for scattcriv• by rain. prcdicts similar results relative to the magnitudes of thescattered iinals. Table I list% resuts foir rain 4cattcr for the two aperture sir•e•s at Ku andtFor the S hours of obwervations in rain with the large aperture antenna. the azerage of thehourly niodian ,ig.nal aels at Ku band .as ) dl louwer than for C hand. F:or thc 216 hoursj,,( ,1,i'n at~on% i rami %tith the small aperture ;mtcana. the Ku hand ignal %;a's 127 dB towerthan the c haCnt signal The ., I dMI ditfe(rcce that resultex fr•om a change in aperture Is w-ithinthe expected nieasuremrew t;ro- of( the 2 dkll value predwted for uauttercr untfomnly fillng

16

References

Anderson, B. T. [1989[; Observations and Analysis of Tropospheric Scatter Propagationduring Rain at 16 GHz and 5 GHz, M. E. Thesis, Thayer School of Engineering,Dartmouth College, Hanover NH.

CCIR [ 1989]; Propagation Data and Predictions Methods Required for Trans-Horizon Radio-Relay Systems, Rept 238-5 (MOD F), Study Group 5 (Propagation in Non-IonizedMecdia), ITU, Geneva.

Crane, R. K. [1981]; A review of transhorizon propagation phenomenA, Radio Sci., 16(E),649-669.

Crane, R. K. [1988]; Troposphe. ic Scatter Propagation at 5 and 15 GHz: Results for a 161km Path, Final Report on Research Conducted Under Subcontract to Geoigia Institute ofTechnology as a Part of the Post-Doctoral Program, Rome Air Development Center,Hanscom AFB, MA.

Hoppe, M. A. [1989]; The Study of the Ku band Tropospheric Scatter Channel during Clear-Air Conditions via Data Collected on the Link between Prospect Mill in Waltham, MAand Mt. Tug in Enfield, NH., M. E. Thesis, Th.yer School of Engineering, DartmouthCollege, Hanover NH.

Matthews, D. [1989]; Internal Memo, The MITRE Corp., Bedford MA.

Parl, S. [1989]; Tactical Ku-band Troposcatter, Internal Memo, Signatron

Rice, P. L., A. G. Longley, K. A. Norton and A. P. Barsis [19651; Transmission LossPredictions for Tropospheric Communication Circuits, Vol. I and II, NBS Tech. Note101, U. S. Dept. of Commerce, Washington D. C.

17

Table 1 Quarter Day Summaries

Ratio of Kuband Cband Ku tanddKuband Cband Cbarnd Kuband Kuband C barndCarrierCarrier Carrer Doppler Doppler DopplerDoppler Delay TransmissionTransmission

Levels Level Level Shift Spread Shift Spread Spread Loss Loss(dB) (dBm) (dBm) (Hz) (Hz) (Hz) (Hz) (nsec) (dB) (dB)

Small Aperture AntennaClear Weather

NightHours 238 238 238 233 238 238 238 186

Average 14 9.-119.6 -104.7 -2.e 9.2 -0 7 2.2 144.7 164.4 149.5Standard Deviation 4.5 4.2 5.7 2.5 2.6 1.1 0.3 38.3

Lognormal m 14.8-119 8.105 0 164 6 149.8

Lognormal s 3.5 5.5Morning

Hours 203 203 203 198 203 203 203 156

Average 15.7 .120.5 -104.8 -2.2 9.2 -0.4 2.1 137.0 165.2 149.5

Standard Deviation 4.1 4.8 5.7 2.7 2.5 0.9 0.3 33.2Lognormal m 15.2 .120.2 -105.0 165.0 149.8

Lognormal s 4.5 5.5Afternoon

Hours 169 169 169 169 169 169 169 136

Average 14.7 -114.4 -998 .2.0 7'.7 -0.2 2.1 127.6 159.2 144.6Standard Deviation 3.8 5.5 6.0 2.5 2.1 1.U 0.3 30.8

Lognormal m 15.2-114.5 -99.3 8.5 138.0 159.3 144.1

Lognormal s 4.0 5 2 5.8 2.5 30.0Evening

Hours 240 240 240 234 240 240 240 179

Average 14.1 -118.1 -104 0 -2.1 8.1 -0.5 2.2 138.1 162.9 148.8

Standard Deviation 39 4 2 5.2 2.2 2.3 1.0 0.3 30.0Lognormal m 14.2 -118.0 -103.8 162.8 148.6Lognormal s 4 0 5.0

RainN-ght

Hours 56 56 56 56 56 56 56 38

Average 12 8.117 0 .104 3 5.4 10.1 C5 2.4 183 6 161.8 149.0

Standard Deviation 6.7 7.5 4.2 6.6 2.9 1.5 0.3 45.3Morning

Hours 39 39 39 39 39 39 39 34

Average 9 6.112 2-102.6 92 9.2 1.0 2.3 173.0 156.9 147.4

Standaid Deviaton 5.3 7 1 6-3 5 9 1 8 2 8 0.4 41.5Afternoon

Hours 5s 58 58 58 58 58 58 45Average 14 1 .114 5 -100 3 7 6 109 1 5 2.4 196.0 159.2 145 1

Stantdard Deviaton 6 3 7 0 7 5 8 2 2 6 2 7 0.3 45.0

Hours 63 61 63 63 63 63 63 50Average 13 3.115 8-102 6 79 98 1 1 23 18802 160.6 1473

Standard Oeviation 5 8 6 2 4 7 5 0 2 6 2 3 0.4 37.6

• •= =18

Table 1 Quarter Day Summaries

Ratio of Ku band C band Ku band Ku band C band C band Ku band Ku band C bandCarrier Carrier Carrie r Doppler Doppler Doppler Doppler Delay TransmissionTransmissionLevels Level Level Shift Spread Shift Spread Spread Loss Loss

(dB) - (dBm) (dBm) (Hz) (Hz) (Hz) (Hz) (nsec) (dB) (dB)Elevated Duct

NightHours 32 32 32 32 32 32 32 27

Average 23.9-115.6 -91.6 -1.4 9.7 .0.2 2.3 126.4 160.4 136.4Standard Deviation 4.6 7.9 10.4 1.0 3.4 0.2 0.1 41.5

MorningHours 23 23 23 23 23 23 23 19

Average 23.9 .115.8 -91.9 -1.2 9.2 -0.2 2.1 143.4 160.6 136.6Standard Deviation 2.6 5.7 6.4 1.5 3.3 0.1 0.0 32.2

AfternoonHours 2 2 2 2 2 2 2 2

AverageStandard Deviation

EveningHours 1 1 1 1 1 1 1 1

AverageStandard Deviation

Large Apeilure AntuinakClear Weather

AfternoonHours 50 50 50 50 50 46

Average 11.4.109.2 -97.8 -0.5 3.8 97.3 154.0 142.6Standard Deviation 2.8 6.0 5.1 1.4 0.3 14.0

lognormal m 11.5-108.9 -97.4 153.7 142.2lognormal s 6.0 5.0

RainAfternoon

Hours 8 8 8 8 8 7Average 9.6 .106.3 -96.7 5.5 6.4 118.8 151.1 141.5

Standard Deviation 2.0 2.8 3.3 7.2 4.2 46.9

19

2.6#Half

-2. 4 0.225* Receive A2.4 . 3Beamuidth

2.2

Ca 2.01 1.,

.) 0 1.6 Half Tr-ansm it 1,00> Beams idths-

S) A 3' at 16 GHzCIO a) 29' at 5 GHz sIleC 29' at 16 GHz

o • 1.2 .2C A C

>0o -. s •

0.63i(r- • . Pr'ospect Mt"Z: 0- Hl sS4 i

-20 0 20 60 100 140 180

Distance from Transmitter (kin)

Figure 1: Troposcatter path profile for a 4/3 effective Earth radius. The scattering volumesare displayed as filled areas. The scattering volume for the 3' antenna at 16 GHzencio~c, s ihe scattering volume for the 29' antenna at 5 GHz.

-- •I II0

871289A.OA7 RECEUMO V!+;NAL MEODIANS.

i-oPe' ct Will to Mt. %r4 !'oooscattor Pooooct Wll to Mt. Tug T?'.escitttr56 . . . .. . . . . . . . . . .I' . . . . . . . . . . . .

12I8 USQACA.OT 6I 712S'O

38

~ * 6.44

-202

LAI

-16 -t46 2 4 6 a 11 '2 14 16 t1 26 22 24 S 2 4 6 16 12 14 14 10 20 22 24

Locol t' s o¶t jr,) Local T 1. l • "urI I

Proooc4t Hill to Mt. Tug 7'ooooit¢tte, Pvo•' Oct Will to Mt. Tug T? oa.u ocotor

9 .7116 . OAT

*MIN

00

-'66 I'

is 1 4 0 *) .- .l t1 0 1 a I

Figure 3: Day plots for the small aperture Ku band transmit antenna under clear weatherconditions on July 12. 1989: time senies of received power deviations fromturbulent forward scatter model predictions, time series of the differencesbetween received power deviations from turbulent forward scatter modelpredictions at Ku and C band. time series and sample cumulative distributionfunction for Ku band two sigma multipath delay spread.

RECEIVEO SIGNAL MEOIANS.

Ku-Car *ior. -12126 OKy-Oeltv * -.Ile.1

P:oeopct Will to Mt. Tug To .ocetts, ProivIct Hill to Mt. Tug T-oosocatter

171269A.OAT

4.

2- -lie

L -120• 2 * -

i -2 1,-13

-e o; -* -1I30:

a 2 6 a Is 12 14 16 Is 29 22 24 a 2 4 1 tio 12 14 ¶• Is 26 22 24Local Ti.., (lMaue) Lseer Time 1lfeOrel

.v poect Will to Mt. Tug Tiogow ttor P...w.tt Hill to Mt. Tug Tretgocetiwr

S.. . . • • -16.,3Sd

A 2I t -,2e

receive p e l

o ii quin4t i ee ifo)6 .4.4.' O C•~lta,.

Figure 4: Day plots for the small aperture Ku band transmit antenna under clear weathercondaitins on July 12. 1989: time series of mean Doppler frequency shift andDoppler frequency spread, time series and sample cumulative distributionfunction for carrier channel and delay channel measurements of Ku bandreceived power levels..

I-

C-..

07120".DAT RECEIYEO SIGNAL MEDIAMS,

K.-BAND MOOULATION. 12 5 MB/S C^OE Kv-BANO. -134.0

WEAT-hER CCNCV'IGNS. CLEAR AIR C-BANO. -112.M

PP*.* t I I0 to Mt. '9 . 0oo c•ltto, ro Wdill to Mt. Tug r~ oo)ociltlo,

*712"O.O Al

~ -1962 .- go-161- . -11 . . . . . . . .. . . .

-121 -126.

_-131 10

S. . K .-b and * _4 . Ku-b &Ad" -t41-

- 1 4 - 0 - 1 4 8 tR. 4-e1,.. . . . ... . . . . . ..

e*#" . . . . . . ... . .. . . .-4 -3 -2 -1 1 1 2 3 4 1 2 4 6 6 11 12 14 I4 '6 24 22 24

Stlndard Ol..itlerl Local TIll. Ho• Il

0IP- octt Will t. Mit "0. I To" octitetoo Proosect Hill it Mt. Tug

Tla.o, lcalttr

-N 4

I0t 26U0* OA AMlq. OATisi

,*qu^q 16,0 \'Q tVII. 1\

Figure 5: Day plots for the large aperture Ku band transmit antenna under clear weatherconditions on July 12, 1989: sample cumulative distribution functions for

received power, received power time series, power spectra for received powerfluctuations and coherency between the fluctuations at Ku and C band.

1-1

P.eeo qct HWll to Mt. Tug Tooeoeatto. 0,*sDoct Hill to Mt. T•Ig Treepot4tter

50 * 204g t F t • 1 . .. 07lT l. 25... .. OA 140 0712"0, CAT 18

18

S14

S... ... .. •12

? 20 *•tlt *IIQ •Qu bejsd U~tq

-10

"-40 Rreceive p e vio Levels- 8C band

-56 I0 2 4 6 0 1$ 12 14 16 1S 21 22 24 S 2 4 6 S If 12 14 14 10 28 22 24

Local Tie Ia. olus Local Ties NIewel

Prospect Hill to Mt. Tug Troosocatter Proseect Will to lMt. Tuo Tr....ictittor

350 67128q.OAT 6712Q5.OAT

251 c 2

vi in

is.is -

IS 43 14 If 151& 6 1 20 22 24 .4 .3 - l S I 2 3 4

Figure 6: Day plots for the large aperture Ku band transmit antenna under clear weatherconditions on July 12, 1989: time series Of received power deviations fromturbulent forward scatter miodel predictions, time series of the differencesbetween received power deviations from turbulent forward scatter modelpredictions at Ku anJ C band, time series and sample cumulative distributionfunction for Ku band two sigma multipadi delay spread.

2S

RECEIYED SIGNAL MEDIANS.Ku-C&,titer, -44.0111Ku-Olay . -106 Of

P.osqe.p t HIll to Pt. T.g T.-•.vo.attst- Prospect Hill to Mt. Tug Trqpiiscatter

e7128". OAT

6

uc 44

-4

-. . -1516

6 2 4 6 8 •11 12 14 16 16 2tl 22 24,0 2 A b, 6 it 12 14 16 18 26 2 24Local Tizes (HouP) LitOa Its i le o)ur"O

lProsloelt Hill to M't. Tug Tropolicatter Pi-oievoct Hillte lt .I Tugl Trepaslcatter"

" 31 1~~7126896.0AT --" 2n -123

I -•136

So 24-* .3 t3 .1,"L",al Ti.es i~ u L.. -4 0i-. 1 l t 10.0

Poec ilt|1 Mti Tl TroplcttgPel e.t Hil t. 0M~it. uoT.p.mt.

Figure "7: Day plots for the large aperture Ku band transmit antenna under clear weatherconditions on July 12, 1989- time series of mean Doppler frequency shift and

Doppler frequency spread, time ieries and sampie cumulative distributionfunction for carrier channel and delay channel measurements of Ku handreceived powec+ levels.

A 26

0) _ 0

CD 0

*~~ u

0 4

x 49~E

XX0

76 0 )

X C. )

xo

V- 0 0

x x i

* E4

U ~66

X 0u

0 n C), 0 t 0 tol aO ~ O~ 0 0 -m -M e. *J Cq

7 - 7 7

E

27

7GT58qA.OAT RECEZYE', SIGNAL qEDIANS.

K,-BAA D ?i. ,JLATtON. 12.5 rlB/S CODE Ku-BANO. -111.S

VEATH-ER CONOITIONS. RAIN C-BAND. -97.N

P,ý-o.. t H ll c t Tug t % o9 o cott.' •'ipo.t kill to Mt. Tug TpoDsctitte,

.71 9A. OAT7 A.AT

!- --

Is - ItO

- -138 13I • - • --b

.... - C-h

cctb -1 (14 jSlaf'lar 001atlfte oca Tlo I Hý*

1 4 P •ooi qc i I I| toll M1 t. T.#l T Pag eciD ter too{ P ~aPO cll llt W ill to Pit. TugI Troga statkaI

g7MA Oar 0

i \

*rSt

r' r Ir2

'l 'r Ira r] r* |

tescuo-t M, Mill1

Figure 9: Day plots for the small aperture Ku band transmit antenna under rainy conditions

on July 5. 1989: sample cumulative distribution functions for received power.received power time series, power spectra for received power fluctuatuons andcoherency between the fluctuatioas at Ku and C band.

28

Posg0oct Hil1l to Pit. ,*go c.c A t.t ~.009tt 141 11 A* t T~g r-aqco o

4 F- 7riA OAT $?flh% OAT39

is

IIt

-211

to -2 C-an 36 1 Q 2 2 4 6 0 Is t2 14 IS IS 20 22 24L~ocal Too Ith..t.1 L oilFe a 1 1441,11s

Piel. til H tI It . r .... 4att. r Pe loott wil ?.j, l.g toa"I.

1, *7Ui•A OhYT * ?0 077 OA

SO W

6 -,,! .4

.3. .. o30

-41. o - o to01%,

* Jil, 4 + • giL ia.. . g .. .. * 4 - - 1

Figure 10: Day plots for the small aperture Ku band transmit antennt3 under rainy conditionson July 5. 1989: tinx series of received power deviations from turbulent forwardscaucr nwoo predictions, time series of the differences between received powaerdeviations fromn turbulent forward scatter model predictions at Ku and C band.time series and sample cumulative distribuilon function for Ku band two sigmiamultipath delay spread.

K2G*e - - o

22-

2to

2 f4

"'2.1

= I.I

44-too

j ~ ~ ~ (i 4 1 4 1 ll. 4 4 o z 1 o t 2 4ý64 1 ." wo Logo ff 14B"Pa

I~l to 4t T~O * t

4 -

'?' k.

Figur 11: Day pkcq for the small aDMrUrT Ku band transmit ant-nna under rwav ony intin July 5. 1989: tune ;cn'cs of nean Doppler frequenc